[
https://issues.apache.org/jira/browse/MAPREDUCE-50?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13204667#comment-13204667
]
Steve Loughran commented on MAPREDUCE-50:
-----------------------------------------
I've seen logs in emails that indicate this is still possible to arise in the
0.20.20x series.
The problem is that the {{ScriptBasedMapping.RawScriptMapping.resolve()} method
behaves as follows
# return "/default-rack" if there is no script
# return null if the script fails
# {{CachedDNSToSwitchMapping.resolve()} (which invokes this inner resolve)
sees the null pointer, doesn't add those entries to its cache, then builds a
list of results for each supplied argument -and for unknown hosts returns null.
# The JobTracker doesn't expect null.
What to do?
* The WARN could somehow be raised to a higher level so that it doesn't get
lost to the user -or the error message changed to be more meaningful to users
who don't want to delve into the source.
* The JT (and its YARN successor) need a policy of handling null-ness.
* Better preflight checking of the shell script
Unit tests could verify how the Namenode, JT and RM handle null entries with a
switch mapping that returns null always; trivial.
> NPE in heartbeat when the configured topology script doesn't exist
> ------------------------------------------------------------------
>
> Key: MAPREDUCE-50
> URL: https://issues.apache.org/jira/browse/MAPREDUCE-50
> Project: Hadoop Map/Reduce
> Issue Type: Bug
> Reporter: Vinod Kumar Vavilapalli
>
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators:
https://issues.apache.org/jira/secure/ContactAdministrators!default.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira
